How do ChatGPT and other AI assistants decide which apps to recommend?

Mostly from how often an app appears in the text they were trained on and, for search-connected assistants like Perplexity, from what editorial roundups and creator videos say. The AI App Discoverability Index (March 2026, 4,265 recommendations across ChatGPT, Claude, Gemini and Perplexity) found no formal predictors such as star rating or review count. Editorial presence and query specificity mattered most.

Does a high App Store rating help an app get recommended by AI?

Not directly. The March 2026 index found no correlation between ratings or review counts and AI recommendations. Ratings still matter once a user opens the store, but the AI shortlist is built from mentions in articles, videos and forums, not from store metadata.

Why does adding the word free change AI app recommendations so much?

Because AI assistants take price qualifiers literally. In the March 2026 index, adding free to a query raised the share of free apps recommended from 31.9% to 70.5%, and only 2 of the top 20 apps overlapped between qualified and unqualified queries. If your app has a free tier, say so wherever your app is described.

What sources does Perplexity cite when recommending apps?

In the March 2026 index, 62.4% of Perplexity's citations for app queries pointed to editorial content such as listicles, reviews and roundups, and 14.2% pointed to YouTube. Getting into a credible best-of article and having creators cover the app on YouTube are the two most direct ways in.

How do creators help an app get cited by AI?

Creator videos are a large share of what AI assistants read and cite. Jellyfish data reported in August 2026 shows YouTube creator content appears in more than 25% of AI assistant responses, and YouTube is 14.2% of Perplexity's app citations. A creator who names the app, states what it does and who it is for produces exactly the kind of source AI systems quote.

How AI Assistants Recommend Apps in 2026 (and How to Get Yours on the Shortlist)

The short version

People increasingly ask an AI assistant which app to use and only then open the App Store or Google Play to check the shortlist. The most rigorous public look at how that shortlist gets built is the AI App Discoverability Index 2026 by Kevser Imirogullari, which sent 195 queries across 15 app categories to ChatGPT, Claude, Gemini and Perplexity on March 19 and 20, 2026 and analyzed 4,265 recommendations. Its findings:

Of 1,230 unique apps recommended, only 16.2% appeared on all four platforms. 54.8% appeared on just one.
The top 100 apps took 38.1% of all mentions. 47.6% of apps were mentioned exactly once.
Ratings and review counts did not predict inclusion. Frequency in training data and editorial presence did.
Adding the word "free" moved the share of free apps from 31.9% to 70.5%.
62.4% of Perplexity's citations were editorial content (listicles, reviews, roundups) and 14.2% were YouTube.
Specific queries surfaced niche apps 65.6% of the time, versus 34.3% for broad queries.

Put simply: AI assistants recommend the apps that other people have written and talked about, in the words those people used. That is a different game from App Store Optimization, and it is winnable by small developers because it rewards being clearly described in the right places, not being biggest.

Why this matters more every quarter

Three data points from the summer of 2026 explain the urgency.

Search is sending fewer clicks. SparkToro's June 9, 2026 analysis of US Google searches found that 68.01% ended without a click in January to April 2026, up from 60.45% in 2024, and that AI Overviews cut click-through rates by nearly 60% where they appear (Search Engine Land, June 2026). AI Mode is still small at 0.34% of searches, but it passed one billion monthly users and its query volume was more than doubling each quarter.

The assistants are enormous. Sensor Tower's State of AI 2026 report (June 16, 2026) notes ChatGPT was the fastest mobile app ever to reach one billion monthly active users, and that ChatGPT, DeepSeek and Gemini together account for nearly 90% of time spent in AI assistant apps (Sensor Tower via PR Newswire).

Creator content is what the assistants read. Jellyfish data reported in August 2026 shows YouTube creator content appears in more than 25% of AI assistant responses (Carusele, August 2026). Combined with the index's finding that YouTube is 14.2% of Perplexity's app citations, creator videos are one of the two largest inputs to an AI app recommendation.

How each assistant builds its shortlist

Assistant	Where the recommendation comes from	What moves the needle	Time to effect
ChatGPT	Training data first; web search when enabled	Being described consistently across many pages over time (entity strength)	Months
Claude	Training data; search when enabled	Same as ChatGPT; niche apps surface on specific queries	Months
Gemini	Google index and Knowledge Graph	Traditional SEO signals plus structured data	Weeks to months
Perplexity	Live web search, always cites	Editorial roundups (62.4% of citations) and YouTube (14.2%)	Days to weeks
Siri and Apple Intelligence (iOS 27)	App Intents your app exposes	Adopting App Intents so Siri can act on your app's content	On your next release

The Apple row is new this year. On June 8, 2026 Apple announced that developers can connect apps to Siri AI capabilities, including personal context understanding, app actions and onscreen awareness, through the App Intents framework, and that Small Business Program developers get access to Apple Foundation Models on Private Cloud Compute at no cloud API cost (Apple Newsroom, June 8, 2026). When Siri can do the job, the app that exposes the job is the one Siri opens.

Six things that actually get an app recommended

1. Be in the roundups

Nearly two thirds of Perplexity's app citations are editorial. Make a list of the ten "best [category] apps" articles that already rank for your job-to-be-done and pitch the authors with a specific reason to include you (a feature they say the category lacks, a free tier, a platform they say is unsupported). One inclusion in a Zapier or NerdWallet style roundup does more than a hundred App Store reviews.

2. Get creators to say the name out loud

Creator videos are both a direct citation source and the raw material assistants learn from. The useful video is not a hype reel; it is a person naming the app, stating the one job it does and who it is for, and showing it doing that job. On IdeaEquity you can run that as a pay-per-install campaign: creators who fit your niche make the videos, you pay per verified install, and the videos keep feeding AI answers long after the campaign ends.

3. Describe the app the way people ask for it

Assistants match query language to source language. If people ask for "a budgeting app that syncs with my bank and has a free plan," your listing, website and press mentions should contain those words, not "financial wellness platform." Check the description with the free App Store audit; it flags copy that names features instead of jobs.

4. Say "free" if it is true

The 38.7 point swing on "free" is the single biggest lever in the index. If you have a free tier, "free" belongs in the first sentence of your listing, your homepage title and every roundup pitch. If you do not, consider whether a genuinely useful free tier is worth more than the conversions you would lose.

5. Own a specific job, not a category

Specific queries returned niche apps 65.6% of the time. "Best habit tracker" is owned by the top 100. "Habit tracker for ADHD that does not use streaks" is winnable. Pick the specific phrasing your best users already use and make it the sentence every source repeats.

6. Make the entity consistent

Same app name, same one-line description, same category words across the App Store, Google Play, your site, your press kit, your social bios, and an llms.txt on your domain. Assistants that are unsure whether two mentions refer to the same app do not count either one.

What not to bother with

Buying reviews. Ratings did not predict recommendations in the index and fake reviews risk the listing.
Keyword-stuffing the description. The Princeton GEO study that most of this field cites found keyword stuffing reduced AI visibility by about 10%.
Waiting for training data to update. ChatGPT and Claude move on a months-long cycle. Perplexity and Gemini move in weeks. Start with the fast ones and let the slow ones catch up.

A 90-day plan

Week 1: write the one-sentence job description and put it everywhere. Add "free" if true.
Weeks 1 to 2: pitch five roundups with a specific inclusion reason.
Weeks 2 to 12: run a creator campaign on the job, not the app. Ask for the name in the first five seconds.
Week 4: adopt App Intents for your top two actions if you are on iOS.
Monthly: ask ChatGPT, Perplexity, Gemini and Claude your ten target queries and log whether you appear. It takes twenty minutes and it is the only scoreboard that matters here.

Want creators making the videos AI assistants cite? Create a campaign on IdeaEquity. Set your price per verified install; creators do the rest.

Sources: AI App Discoverability Index 2026, Kevser Imirogullari, March 2026; Search Engine Land on SparkToro's zero-click study, June 2026; Sensor Tower State of AI 2026, June 16, 2026; Apple Newsroom, June 8, 2026; Carusele, August 2026.